Академический Документы
Профессиональный Документы
Культура Документы
*
Author to whom correspondence should be addressed.
Abstract
The analysis of the feasibility for acquiring automated systems requires the
engineer to estimate the future costs and benefits through forecasts and/or predictions.
The uncertainty of the future makes the engineer consider a set of alternative capital
investments so that the best solution to the problem can be obtained and implemented.
The best solution can be obtained by evaluating the active life cycle cost of the
automated system to be acquired. System active life cycle includes procurement,
operation, maintenance, and retirement (phase-out) activities. The objective of this paper
is to, (i) propose a breakdown structure identifying the main activities and sub-activities
for the active life cycle of automated systems, and (ii) construct a discounted cost model
to calculate the total cost for the active life cycle of automated systems. Comparison
between the results obtained using the proposed cost model and the real cost for the
system active life cycle shows a difference error of fifteen percent.
1.1 Background
Investments are distributed over a period of time, and time has a monetary
value. These investments include machinery, equipment, facility, and so on. The analysis
of the feasibility of these investments requires the engineer to estimate the future costs
and benefits through forecasts and/or predictions. The uncertainty of the future makes
the engineer consider a set of alternative capital investments, so that the best solution to
the problem can be obtained and implemented. The best solution will answer questions
as: best investment alternative (e.g., buy a new equipment or repair the old one, degree
of automation versus labor employed, and so on).
Determining the worth of any capital investment alternative, several variables
must be defined. In general managers consider the following variables in the decision
making process of capital investment: (i) system total cost; (ii) expenditure and
benefits/revenues, and the time when they will occur; (iii) the economic life of the
alternative (life cycle planning); (iv) and the satisfaction with the investment. The
definition of most of these variables is accomplished by identifying and evaluating the
cost for the activities of the system active life cycle. The system active life cycle starts
after the system leaves the supplier’s facility, and ends when the system retires. System
active life cycle includes system procurement activity, system operation activity, system
maintenance activity, and system retirement (phase-out) activity. Gaimon (1985)
proposes an optimization model focusing on the continuous acquisition of automation
that is used to enhance the productivity of an organization’s existing workforce. Hegde
(1994) proposes a life cycle cost model to estimate the life cycle of durable products.
Their cost models are very general, and do not identify the main active life cycle
activities, and how to calculate their costs.
1.2 Objective
2
equipment, maintenance, and so on. These two approaches are usually used in different
stages of the system purchase planning, where the top-down approach is used in the
early stage and the bottoms-up approach is used in the later stage. Both approaches
require the identification of several activities, that per se are related to several cost
categories.
In this study, a breakdown structure is presented identifying the main activities
for the active life cycle activities of automated systems−Chapter 2. Chapter 3 proposes a
discounted cost model to calculate the cost for each activity of the breakdown structure,
that is presented in Chapter 2. Chapter 4 presents a numerical example, and Chapter 5
the conclusion of this study.
Estimating the cost for the active life cycle of automated systems (AS) is a very
complex task, that includes the cost evaluation of several activities. Figure 1 depicts a
three level breakdown structure for the active life cycle of AS. All activities of the
breakdown structure have two subscripts. The first subscript indicates the activity level,
and the second subscript is the activity number. For example, the activity X2,3 is at level
two, and its identification number is 3. Each level 3 activity may include the following
cost categories: labor cost, utility, support equipment, consumable material, support
tools, material/equipment for personnel, facility, and computer software.
3
Active Life Cycle
Installation (X2,2) Operation actions (Y2,2) Maintenance actions (Z2,2) Retirement actions (W2,2)
- facility preparation (X3,11) - setup the AS (Y3,11) - maintain support equip. (Z3,11) - disassemble the AS (W3,11)
- unpack system (X3,12) - check the AS (Y3,12) - preventive maintenance (Z3,12) - recycle (W3,12)
- assemble system (X3,13) - start, adjust, operate AS (Y3,13) - corrective maintenance (Z3,13) - disposal (W3,13)
- perform overhaul (Z3,14)
Initial inspection (X2,3) Personnel training (Y2,3) Personnel training (Z2,3) Personnel training (W2,3)
- check AS (X3,21) - literature preparation (Y3,21) - literature preparation (Z3,21) -literature preparation (W3,21)
- check environment (X3,22) - seminars (Y3,22) - seminars (Z3,22) -seminars (W3,22)
- check support equipment (X3,23)
Personnel training (X2,5) Safety (Y2,5) Safety (Z2,5) Tech. data prep. (W2,5)
- literature preparation (X3,41) - inspections (Y3,41) - inspections (Z3,41) - data collection (W3,41)
- seminars (X3,42) - information prep. (Y3,42) - information prep. (Z3,42) - data update (W3,42)
- data collection/prep. (Y3,43) - data collection/prep. (Z3,43)
Safety (X2,6) Technical data preparation (Y2,6) Technical data preparation (Z2,6)
- inspections (X3,51) - data collection (Y3,51) - data collection (Z3,51)
- information prep. (X3,52) - data update (Y3,52) - data update (Z3,52)
- data collection/prep. (X3,53)
4
technicians, non-technical workers, and so on). Personnel benefits are also included (e.g.,
bonus, holidays, sickdays, social security, health insurance, legal costs, vacation,
company contribution of profit sharing, and so on). The labor cost category can be
calculated by
n L m p
∑( )( ) ( )( )
L4,0 = L5,1 ⋅ L5,2 + L5,3 + L5, 4 ⋅ L5,5 + L5,6 + 5, 2 ⋅
L5, 7
∑( L ) 5, 8 j + ∑( L ) 5, 9 k ⋅e
ρη
(1)
i =1 j =1 i k =1
n ρη
∑[ ( ) ] + ∑[ ( U ) ⋅ ( U ) ]
m
U 4, 0 =
i =1
)(
U5,1 ⋅ U 5, 2
i
5, 3 5, 4
j
⋅e (2)
j =1
ρη
∑[ ( ) ] + ∑[ ( E ) ⋅ ( E ) ⋅ ( E ) ]
n m
E4, 0 =
i =1
) ( ) (
E5,1 ⋅ E5,2 ⋅ E5,3
i
5, 4 5, 5 5, 6
j
⋅e
(3)
j =1
5
i = 1,2,3,…, n, where n is the number of different rented equipment
j = 1,2,3,…, m, where m is the number of different purchased equipment
The term E5,2 is the total rental time for the equipment, which is different from
the total usage time for all activities (Tt). The reason is that sometimes support
equipment may not be in use ( E5,2 ≤ Tt ), and still be under the user’s possession.
ρη
∑[ ( ) ] + ∑[ ( M ) ⋅ ( M ) ] + ∑ ( M )
n m p
M 4, 0 =
i =1
)(
M5,1 ⋅ M5, 2
i
5, 3 5, 4
j
5, 5 k ⋅e
(4)
j =1 k =1
The second term of Equation (4) should be included only when calculating the
cost for the inventory activity (Z2,8).
n n
T4, 0 = ∑[
i =1
] ∑ [ (T
(T5,1 ) ⋅ (T5, 2 ) ⋅ (T5,3 ) i +
j =1
5, 4 ]
) ⋅ (T5,5 ) ⋅ (T5, 6 ) j ⋅ e ρη
(5)
6
equipment, or any kind of device that is used/worn by any personnel to perform a
particular task. The equipment considered in this category are non-depreciable
equipment. Material/equipment cost can be calculated by
n m
P4,0 = ∑ ( P5,1 ) ⋅( P5, 2 ) + ∑ ( P5,3 ) ⋅( P5, 4 ) ⋅ e ρη
i =1 j =i
(6)
∑{ [ ( F ) ⋅ ( F ) + ( F ) + ( F ) + ( F ) ] ⋅ ( F )} + ∑{ [ ( F ) + ( F ) + ( F ) ] ⋅ ( F )}
n m ρη
F4, 0 = 5,1 5, 2 5, 3 5, 4 5, 5 5, 6 5, 7 5, 8 5, 9 5,10 ⋅e (7)
i = i
j =1
j
i = 1,2,3,…, n, where n is the number of different purchased facilities
j = 1,2,3,…, m, where m is the number of different rented facilities
n ρη
∑[ ( ) ( ) ( ) ( ) ( ) ] + ∑[ ( S ) ⋅ ( S ) ⋅ ( S ) + ( S ) ]
m
S4, 0 = S5,1 ⋅ S5,2 ⋅ S5,3 ⋅ S5, 4 + S5, 5 5, 6 5, 7 5, 8 5, 9 ⋅e (8)
i =1 i
j =1
j
7
3.2 Cost Model Formulation
In words, Equation (9) means that all workstations are disjoint (mutually exclusive). For
any number of disjoint events (workstation failures),
P ( S ) = P ( w1 ∪ w2 ∪...∪ wn ) . (10)
Then,
P( S ) = P ( w1 ) + P ( w2 ) +...+ P ( wn ) (11)
8
P(w2) = probability of failure for workstation #2
10. One of the main objectives of the proposed model is to obtain the system up time
(operation time) and the down time (preventive maintenance actions, corrective
maintenance actions, and when overhauls are needed), to calculate the cost to operate
and maintain the system. It is assumed that data from a similar system is available.
11. Preventive maintenance actions depend on the number of operating hours. For
example, for every one hundred hours of operating time, preventive maintenance type I
must be performed, for each two hundred hours of operating time preventive
maintenance type II is performed, and so on.
12. As the system ages, the failure rate λ increases, due to wear out of components, and
so on. So there is a point in time that the system failure rate will reach a high value λMax,
and a normal preventive maintenance action will not be enough to make the system
operate efficiently. So a more detailed preventive maintenance (overhaul) is required.
The overhaul Ov is required periodically during the system life cycle and it can be also
planned in advance.
9
see Kapur & Lamberson (1977). Figure 2 depicts the life cycle of AS. The system starts
with an initial failure rate λi , which can be determined by using data from a similar
system, testing, and/or computer simulation. After certain time the failure rate reaches a
high value λMax. At this time tOv overhaul is performed. During the system life cycle
several overhauls will be required to maintain the system operating efficiently. As the
system ages, the time between two overhauls gets smaller. At time tR, an overhaul will
not bring back the system to an efficient operating status. This is the retirement time or
phase-out.
The system life cycle is composed of n renewal cycles ρ, periodic preventive
maintenance and overhaul. The life cycle is given by
( )
t R = α + β ⋅ n + t p + tOv (12)
λMax( R )
r(t) λMax λMax λMax
λi
Figure 2. Hazard function r(t) over the system life cycle. λ = r(t)
Table 1 shows the cost category models that apply to each active life cycle
activity, and the activities that are time dependent. There are three activities that are time
dependent: operation action activity Y2,2, that depends on operation and maintenance
time; corrective maintenance action Z2,2, and inventory Z2,4, that depend on maintenance
time. The variables for each time dependent activity is given bellow:
10
Operation action activity (Y2,2): L5,2, L5,3, U5,2, M5,2, M5,4.
Corrective maintenance activity (Z2,2): L5,2, L5,3, U5,2, E5,4, E5,7, M5,2, M5,4, P5,1, P5,3.
Inventory− spare/repair parts (Z2,3): L5,2, L5,3, U5,2, E5,4, E5,7, F5,6, F5,10.
The cost for these activities is obtained by first determining the expected operational
time, expected maintenance time, preventive maintenance time, and overhaul time.
Second, a relationship between each variable and the expected times must be realized.
For example, (i) U5, 2 = b ⋅ α (t), where b is the amount of consumed energy (utility type
i) per operating and/or maintenance hour (e.g., m3/hour); (ii) L5, 2 = a ⋅ α (t), where a is
the proportional labor rate, that determines the percentage of labor hours for all labor
types−for example, if α = 100 hours and 5 hours goes for management, 10 hours goes
for supervision, and 85 hours goes for operation, then am = 0.5 , as = 0.15 , and
ao = 0.85 ; and so on. Third, plug back the variables’ values into their respective cost
category models. Finally, calculate the activity cost adding up the cost category models
that should be considered, according to Table 1 (e.g., Y2,5 = U4,0 + E4,0 + T4,0 + P4,0 + F4,0 +
S4,0).
If data from a similar system is available (testing, and/or computer simulation)
density functions for α and β can be obtained, and finally their respective expected
values Eα and Eβ. It is desired to obtain the expected values for each year, to compute
the annual cost for the active life cycle of AS.
where u( t ) wi is the density function of operating time (time to failure) for ith
workstation, and t[l] , t[u] are the lower and upper bound, respectively.
The expected value for α is given by
n ∞
Eα ( system ) ( T ) = ∑ ∫ t ⋅ u(t ) dt , i = ,2,3, … , n (14)
i =1 0 i
11
operating hours ( t [ l ] = 0, t [ u ] = 12 months ).
(15)
12
where p = elapsed time for preventive maintenance type i
mp = number of preventive maintenance type i (year basis)
P( α ∩ β )
P( β α ) ≡ , P(α) > 0 (16)
P( α )
The random variables α and β are strictly continuous. The conditional density function of
β given α is
f ( α, β)
f ( β α) ≡
f1 ( α )
(17)
tβ [ u ]
n
f (α,β)
f ( β α ) system ≡ ∑ t [u ] , i = 1,2,3,…, n (18)
i =1
β
∫ f1 ( α , β ) dβ
t β [ l ] i
In Equation (18), f(α,β) is the joint function of α and β for workstation i, and
tβ [ u ]
tβ [ u ]
E( β α ) = ∫ β ⋅ f ( β α) dβ
tβ [ l ]
(19)
13
where α is to be interpreted as αl < α < αu (αl = lower bound, αu = αl + dα = upper
bound).
Finally the expected value for the corrective maintenance, for the entire system,
is given by
n tα tβ
[ u]
[u]
Eβ ( system ) (T ) = ∑ ∫ ∫ β ⋅ f ( β α ) dβ ⋅ f1 ( α ) dα
i =1 tα [ l ] t [ l ]
β i (20)
( )
n
Ov ( system ) = ∑ mOv ⋅ Ov
i
i =1
(21)
4. Numerical Example
14
match the buyer’s specifications and requirements. The criteria to choose the best system
are: (i) lowest active life cycle cost, that includes system acquisition cost, procurement
cost, operation cost, maintenance cost, and retirement cost; (ii) lowest system total
downtime; (iii) and highest production rate.
A utility function U ( x ) −see Keeney and Raifa (1993)−will be used to select the
best system, which is the one with the highest utility value.
Table 2. Annual (end of year) active life cycle cost for system A
Active Begin. 1st 2nd 3rd 4th 5th
Life Cycle of 1st y year year year year year
Initial cost 4000000 0 0 0 0 0
Procurement 0 11849 0 0 0 0
Operation 0 99996 100331 100386 100307 99847
Maintenance 0 74587 74732 75013 75328 75193
Retirement 0 0 0 0 0 46415
Salvage Value 0 0 0 0 0 -1000000
TOTAL 4000000 186432 175063 175399 175635 -778545
The life cycle data for system B was provided by a supplier, from USA, that
uses computer software to simulate the system life cycle of AS. Table 3 shows the annual
cost (end of year ) for the active life cycle of system B (values are not discounted).
Table 3. Annual (end of year) active life cycle cost for system B
Active Life Begin. 1st 2nd 3rd 4th 5th
Cycle of 1st y year year year year year
Initial cost 3700000 0 0 0 0 0
Procurement 0 8500 0 0 0 0
Operation 0 112500 113300 111779 110128 109628
Maintenance 0 85876 86432 88987 90765 89978
Retirement 0 0 0 0 0 95780
Salvage Value 0 0 0 0 0 -520000
TOTAL 3700000 206876 199732 200766 200893 -224614
The life cycle data for system C was provided by a third supplier, also from
USA, that uses the same method that supplier of system B uses (computer simulation) to
generate data for the system life cycle. Table 4 shows the annual cost (end of year ) for
the active life cycle of system C (values are not discounted).
15
Table 4. Annual (end of year) active life cycle cost for system C
Active Life Begin. 1st 2nd 3rd 4th 5th
Cycle of 1st y year year year year year
Initial cost 3200000 0 0 0 0 0
Procurement 0 6200 0 0 0 0
Operation 0 115890 117146 118685 116788 117890
Maintenance 0 93987 94059 96986 98890 97002
Retirement 0 0 0 0 0 75900
Salvage Value 0 0 0 0 0 -600000
TOTAL 3200000 216077 211205 215671 215678 -309208
[
Ui ( x ) = ∑ ( w1 ⋅ x1 ) + ( w2 ⋅ x2 ) + ( w3 ⋅ x3 ) + + ( wn ⋅ xn )
j =1
] (22)
j
16
where Ui ( x ) = total utility value for system i, 0 < Ui ( x ) ≤ 1
x = normalized attribute vector x1, x2, x3, … , xn
wj = normalized weight value for attribute xj
6000000
5755896
5500000 5,257,265 (B)
5271119 5427775 5,051,335 (A)
Discounted Cost ($)
5000000
4820197 4971964 4,901,211 (C)
4883111
4538504
4500000
4401075 4394376
4000000 4107186
3974907 System A
3700000 System B
3580145
3500000
System C
3200000
3000000
0 1 2 3 4 5
Life Cycle (years)
Figure 3. Plot for the accumulative annual cost for the active life cycle of systems
A, B & C.
The systems’ attributes considered in this study are: (i) system accumulative
cost, x1 (total life cycle cost); (ii) system up time, x2; and (iii) high quality operation, x3.
Table 5 shows the values for the utility analysis of systems A, B, and C. The last
column shows the total utility value for system A, B, and C (0.36, 0.33, and 0.31
respectively). System A has the highest total utility value.
The objective is not only to choose the system with lowest cost, but to choose
the system that will satisfy all criteria (i.e., attributes x1, x2 , and x3) simultaneously.
Although system A has the highest life cycle cost, its total utility value is the highest.
Thus, the final decision is to choose system A. The discounted cost model, proposed in
Chapter 3, was used to evaluate the total cost for the active life cycle of systems A, B,
and C.
Using real data for the active life cycle of each system, an error analysis was
performed, and the percentage error was found to be 15 percent.
17
Table 5. Utility analysis values for systems A, B, and C
System x1 w1 x2 w2 x3 w3 Ui
A 0.33 0.5 0.37 0.2 0.40 0.3 0.36
B 0.32 0.5 0.34 0.2 0.35 0.3 0.33
C 0.35 0.5 0.29 0.2 0.25 0.3 0.31
5. Conclusion
Many cost models have been developed in the last decade, and most of them
consider only operation and maintenance activities, disregarding other main activities and
their sub-activities. Those cost models are very general and do not allow one to define
specifically the cost for an activity alone. Hegde (1994) proposes a life cycle cost model.
The objective of his study is towards estimating life cycle cost of automated systems
considering discounting issues and nonlinear costs of product failures, and not to
investigate, which he calls, the variable cost C(t) −that includes operation and
maintenance. Gaimon (1985) also does not investigate the active life cycle, but the
relation between acquiring AS and reduction of labor. The same does Houtzeel (1981)
and Tombari (1978).
In this paper a cost analysis was proposed with the intent to support AS
purchase strategy. The proposed model allows one to evaluate the cost of any active life
cycle activity and/or sub-activity, at any level. The necessary data, to apply the proposed
cost model, can be obtained by a previous similar system, testing, and/or computer
simulation. The proposed cost model have not been compared with other cost models.
Actually, real data, from three different companies, were used to check the results that
were obtained by applying the proposed model. The results differed by 15 percent
(existing models have an average percentage error that ranges from 10 to 20 percent).
The percentage error can still be reduced if more activity levels are added to the
proposed breakdown structure in Figure 1.
Next step is to develop a windows oriented software (e.g., Visual Basic, Visual
C++, etc.), that will reduce the data input time, and allow one to reuse the current cost
analysis for different AS. This task is underway.
Acknowledgments
This work was made possible by Kyoto University and by the Japanese Ministry
of Education (Mombusho).
18
Appendix
Nomenclature
E5,1 = rent cost per unit of time ($/hour) for equipment type i
E5,2 = total rental time for equipment type i in one year (all activities)
E5,3 = allocation factor for rented equipment type i. The allocation factors E5,3 and E5,6,
are used when the support equipment/tool can be used in different activities (e.g.,
a crane can be used during system operation and also during system maintenance
action). The allocation factor value can be determined according to the time of
use in each activity. For example, if an equipment is used in the operation activity
only, the allocation factor should be 1.0; if the equipment is used 100 hours in the
operation activity and 30 hours in the maintenance activity, during the current
year, then the allocation factor should be 0.77 for operation activity and 0.23 for
maintenance activity.
E5,4 = cost per unit of equipment j (equivalent annual cost). This value should be
allocated throughout the support equipment life cycle. It should be equivalent to
the cost of using the equipment in one year (equivalent annual cost).
E5,5 = total number of purchased equipment type j (in one year)
E5,6 = allocation factor for purchased equipment type j (see E5,3)
F5,1 = rate cost per square meter of land/building for facility i (equivalent annual cost)
F5,2 = total amount, in m2, for facility i
F5,3 = cost for building facility i (equivalent annual cost)
F5,4 = cost for maintaining facility i (equivalent annual cost)
F5,5 = cost for restoration/expansion of facility i (equivalent annual cost)
F5,6 = allocation factor for using facility i
F5,7 = rent cost for facility i (equivalent annual cost)
F5,8 = cost for restoration/expansion of facility i (equivalent annual cost)
F5,9 = cost for maintaining rented facility i (equivalent annual cost)
F5,10 = allocation factor for using rented facility i
19
system failure/maintenance actions). This term applies only to personnel that
operates the AS.
L5,4 = overtime rate cost ($/hour) for personnel type i
L5,5 = overtime labor in hours for personnel type i (overtime for the current activity in
one year.
L5,6 = idle overtime time, due to system failure, for personnel type i (during this time the
personnel that operates the system do not perform any task due to full/partial
system failure/maintenance actions). This term applies only to personnel that
operates the AS.
L5,7 = the total number of worked regular hours in one year for personnel type i (usually
1920 hours, 40 hours per week) per worker.
L5,8 = cost of benefit type j per year, for personnel type i
L5,9 = extra cost (it includes all extra costs not covered anywhere else)
M5,1 = cost rate per unit of material i (e.g., $/m3, $/m2, $/m, $/liter, $/Kg, etc.)
M5,2 = quantity of material i needed per year (e.g., m3, m2, m, liter, kg, etc.)
M5,3 = cost rate per unit of spare/repair part type j
M5,4 = number of spare/repair part type j
M5,5 = extra cost, which includes cost for insurance, taxes, obsolescence, and spoilage
of loss.
S5,1 = cost type I: lease cost per hour per computer for software type i
($/hour/computer); cost type II: lease cost per hour for software type i.
S5,2 = total lease time for software type i (total time in one year)
S5,3 = number of computers with software type i installed (only for cost type I)
S5,4 = allocation factor for using software type i (see E5,3 Equation 3)
S5,5 = cost of software support for software type i (for one year)
S5,6 = cost for one software type j (equivalent annual cost)
S5,7 = number of computers with software type j installed (only for cost type I)
S5,8 = allocation factor for using software type j (see E5,3 Equation 3)
S5,9 = cost of software support for software type j (for one year)
20
T5,1 = rent cost per unit of time ($/hour) for support tool i
T5,2 = total rental time for support tool i (in one year)
T5,3 = allocation factor for rented support tool i (see E5,3 Equation 3).
T5,4 = cost per unit of support tool j
T5,5 = total number of purchased support tool j (in one year)
T5,6 = allocation factor for purchased support tool j (see E5,3 Equation 3)
U5,1 = cost rate for utility type i, e.g., $/kWh (electric bill), $/m3 (gas), etc.
U5,2 = amount of utility type i spent in one year
U5,3 = property tax type j (total for the current year)
U5,4 = allocation factor, which is equal to F5,6 and/or F5,10 (Equation 7)
References
Gaimon, C., 1985, The Optimal Acquisition of Automation to Enhance the Productivity
of Labor, Management Science, 31, 1175−1190.
Hegde, G. G., 1994, Life Cycle Cost: A model and Applications, IIE Transactions, 26,
56−62.
Haviland, R. P., Engineering Reliability and Long Life Design. Princeton: D. Van
Nostrand Company, Inc., 1964.
Jarret, J., 1991, Business Forecasting Methods (Basil Blackwell, Inc., Cambridge).
Kapur, K.C. and Lamberson, L. R., 1997, Reliability in Engineering Design (John Wiley
& Sons, Inc., NY).
Keeney, R. L. and Raifa H., 1993, Decision with Multiple Objects−Preferences and
Values Tradeoffs, Cambridge University Press.
21
Mechanical Engineering, Graduate School of Engineering, Kyoto University,
Japan.
Yoshimura, M. and Tekeuchi A., 1994, Concurrent Optimization of Product Design and
Manufacturing Based on Information of User’s Needs, Concurrent Engineering:
Research and Applications. Vol. 2, pp. 33-44.
22